Assimilation of satellite data for the environment · Summary of introduction {Use of satellite...

ECMWF seminar ‘07 F. Chevallier

Assimilation of satellite data for the environment

Frédéric Chevallier, Peter RaynerLaboratoire des Sciences du Climat et de l’EnvironnementCEA/CNRS/UVSQ, IPSL, France

Richard EngelenECMWF, Reading

ECMWF seminar ‘07 F. Chevallier slide 2

Thanks

Audrey Fortems, Philippe Peylin and Sophie Szopa

LSCE

The members of the FP6 GEMS consortiumTony HollingsworthAntje Dethof and Angela Benedetti


Outline

Introduction: CO2 from spaceOptimization of atmospheric concentrationsOptimization of surface fluxesOptimization of surface model parametersConclusion


The atmosphere from space

IASI Level 1C Spectra 29/11/2006, 13:42:11 UTCSource CNES-CNRS ETHER

aerosols aerosolsaerosols


CO2 as noise

HIRS 15-micron channels

“The brightness temperature differences can be as large as 1 K for a 30-ppmv CO2 increase and a seasonal variation of a few tenths of a Kelvin may exist” (Turner, 1993).Impact on temperature retrievals

Turner, 1993

Tb(CO2 variable) –Tb(CO2@330ppm) [K]

CO2 [ppm]


CO2 as signal

First retrievals of CO2 concentrations (Chédin et al., 2003).Upper troposphere in tropical latitudesHIRS+MSU1987-1991Neural network

Model peak to peaktwice as less

Chédin et al., 2003

344 ppm 368344 ppm 368


CO2 as signal (cont’ed)

Extension to high-spectral resolution measurementsAIRSUpper troposphere in tropical latitudes

Engelen et al., 20041D-Var

370 ppm 384372 ppm 381

Crevoisier et al., 2004Neural networkAugust 2003


CO2 as signal (cont’ed)

Extension to high-spectral resolution measurementsSCIAMACHYTotal column over lands

Buchwitz et al., 2005DOAS

Model simulations of CO2column [ppm]Olsen and Randerson, 2004


The same planet?

Total column vs. upper tropospheric columnRetrieved vs. measured

Barkley et al., 2006SCIAMACHY (red) andAIRS (blue) over North America

In situ measurements of CO2mixing ratios in the free troposphere (FT) and in the continental boundary layer (CBL), and of NEE (Hurwitz et al. 2004).


CO2 as primary target

CO2 concentrations higher than at any time within the last 650,000 years

OCO (NASA)Launch Dec’ 2008

GOSAT (JAXA, NIES, MoE)Launch Dec’ 2008

More projectsA-SCOPE, ACCLAIM,

CARBOSAT, …

IPCC (2007)


Beyond NWP within NWP systemsErrors on atmospheric concentrations of gases and aerosols affect NWP systems

NWP systems flexible and powerful enough to tackle environmental issues

Expertise in data mergingExpertise in satellite observationsExpertise in atmospheric modelling

Surface and soil properties

NOAA-10 channel 10Calculated minus obsPierangelo et al., 2004


FP6 GEMS project

Part of the Global Monitoring for Environment and Security (GMES, funded by EC & ESA) Atmosphere theme

31 consortium members, 4 years (started in March 2005)

Coordinated by ECMWF

Creation of a pre-operational global monitoring system for greenhouse gases, reactive gases, and aerosols in the troposphere and in the stratosphere

Near-real-time and retrospective global analyses for monitoring atmospheric composition, and short-range forecasts to drive regional air-quality models.


Summary of introduction

Use of satellite data for the environment = emerging topicIncreasing interest from the NWP communitySignal-to-noise may be challenging for some species

MODIS/Terra, August 2003, daytimeAerosol optical depth @0.55 micron

MISR/Terra, August 2003Aerosol optical depth @0.55 micron

A. Benedetti


Outline

IntroductionOptimization of atmospheric concentrationsOptimization of surface fluxesOptimization of surface model parametersConclusion


Mathematical framework

Bayes’ theorem

The optimal solution minimizes the following cost function

x: state vector (atmospheric concentrations)y = Hx+ε : observation (atmospheric concentrations)H: linear observation operator

(short-range chemical transport + interpolation)B: background error covariance matrixR: observation error covariance matrix


CO2 analysis

Analysed CO2

Analysis minus free runAugust 2003

CO2 4D-Var analysis using AIRSAugust 2003Started in January 2003

R. Engelen


Individual Profiles

10 ppmv

11 May 2003

Comparisons with flight data over Hawaii (courtesy of Pieter Tans, NOAA/ESRL) shows a clear improvement of the analysis over the free-running model.But, background error is important.

25 September 2003R. Engelen


Individual Profiles

10 ppmv

11 May 2003


25 September 2003R. Engelen


Individual Profiles

10 ppmv

11 May 2003


25 September 2003


Molokai, Island, Hawaii

Blue: free-running model

Red: reanalysis

Black: observationsR. Engelen


Analysis of CO

CO2 lifetime ~ 100 yearsCO lifetime ~ 2 monthsCO interacts with OH

Surface sources (combustion)Chemical production in the atmosphereChemical loss in the atmosphere

Observed by MOPITT satellite since 2000

GEMS analysis system : 2-way coupling between IFS and a chemistry-transport model


Analysis of COAssimilation of MOPITT dataFree running

15-30 July 20031018 modelcules/cm2

Assimilation minus free run

A. Dethof


Analysis of CO (cont’ed)

Assimilation of MOPITT CO columns leads to improved fit to profile observations from MOZAIC flights

15-30 July 2003MOSAIC data over Osaka, JPN

MOZAIC flight

Free-running CO model

Assimilation of MOPITT data

A. Dethof


Assimilation of POLDER data within the LMDZ-INCA model

Aerosol optical thickness

AN-FG AN-FG POLDER

Generoso et al. (2007)



AN-FG AN-FG POLDER



Outline




Bayes’ theorem


x: state vector (surface fluxes)y = Hx+ε : observation (atmospheric concentrations)H: linear observation operator

(long-range chemical transport + interpolation)B: background error covariance matrixR: observation error covariance matrix


CO emissions from MOPITT

Comparison of a priori (grey symbols) and a posteriori (black symbols) monthly biomass burning sources in Africa with van der Werf et al. (2004) inventory (white symbols)

Petron et al. (2004)


CO2 fluxes from AIRS

11-month inversionMarch 2003, GEMS test re-analysisAN-FG, gC/m2/month

Inversion from satellite data

Inversion from in situ data(P. Peylin, pers. comm.)


CO2 fluxes from AIRS

11-month inversionMay 2003 , GEMS test re-analysisAN-FG, gC/m2/month

Inversion from satellite data

Inversion from in situ data(P. Peylin, pers. comm.)


Multi-tracer inversions

Simplified atmospheric chemistryComputing timeLimited observation information content

Hydrocarbon oxidation chain


Multi-tracer inversion from MOPITT+ surface MCF

Mean January 2007MOPITT L2V5.9.4700 hPa (ppb)31,400 retrievals

First-Guess equivalent Analysis equivalent


Multi-tracer inversion from MOPITT+ surface MCF

CO flux increments(g/m2 for 1 month)

HCHO-productionscaling factor (0-1)


Independent HCHO observation

Too much HCHO in the free model

F. Wittrock, 2006Bremen University


Outline




Bayes’ theorem


x: state vector (model parameters)y = Hx+ε : observation (atmospheric concentrations + …)H: linear observation operator

(surface model +long-range chemical transport + interpolation)B: background error covariance matrixR: observation error covariance matrix


Con: involving development

More complex observation operator

LMDZT transport model includes ~ a few thousands lines of codeORCHIDEE model of the terrestrial vegetation includes ~ 40,000 lines of code

1 1( ) 2 ( ) 2 ( [ ])TJ H− −∇ = +bx B x - x H R y - x


Con: model as hard constraint

Changes the observation errors as seen by the inversion system

[Observation error] = [Measurement error] + [representativeness error] + [Model error]

Biases / Variances / Correlations

We may not have enough information from the observations to introduce a weak constraint formulation


Impact of observations error correlations

Surface fluxes from OCOHypothetical O.5 along-track correlation Correlations ignored in the inversionUncertainty reduction 1-sig(post)/sig(prior)

Increment in error reductionReference error reduction

Chevallier (2007)


Pro: assimilate more than atmospheric concentrations

020406080

100120140160

Free Run MODIS-Assimilated Run

evergreen needle-leafforestevergreen broad-leaved forestdecidious broad-levedforestmixed forest

natural C3

agricultural C3Demarty et al. (2007)

Assimilation of MODIS LAI within the ORCHIDEE vegetation modelRMS difference between simulated gross primary production and independent FLUXNET data (40 sites)

gC/m2/month


Pro: spread increments

Background error correlations

CO2 fluxes:Prior error spatialcorrelationsat a series of sitesChevallier et al. (2006)


Pro: spread increments (cont’ed)

Optimizing generic parameters may be more efficient than prior errors in spreading the observation information in space and time

Error reductionfor the inversion of CO2surface fluxes from CO2concentrations at two sites. 4-day period.

No prior error spatial correlations.

Transport from MesoNHat 8-km resolution.

Lauvaux et al. (2007)0.1 0.9


Pro: predictive capability

Rayner et al. (2005, 2007)


Summary

From the assimilation of satellite data to the inversion of parametersComprehensive approachIncreased sophisticationLarge networks of expertise required


2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

NASAEurNOAAJAXA

IASI

DOAS Sounders

GREENHOUSE Gases : Main Satellite Provision 2003-2019

Advanced Sounders

AIRS

ENVISAT (SCIAMACHY) Uncertainty

GOSAT

NPP/ CrIS

NPOESS / CrIS

OCO

Greenhouse gas provision


Reactive gas provision

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

NASAEurNOAAJAXA

AURA (TES, OMI)

AURA (TES, OMI)

ENVISAT (MIPAS, SCIAMACHY, GOMOS)

Lower Troposphere

ENVISAT (SCIAMACHY)

NPP/ OMPS (~sbuv+toms)

OMPS-Nadir (Npoess)

GOME (Metop)

REACTIVE Gases (O3, N2O, SO2, CH2O) : Main Satellite Provision 2003-2019

Upr. Trop. -Lower StratUncertainty


Aerosol provision

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

NASAEurNOAAJAXA

V-IR Imager

V-IR Imager

Uncertainty

NPOESS/ VIIRS

MERIS

AURA (TES, OMI)

APS (Glory)

ParasolPolarimeter

AEROSOLS, Albedo, Ocean Colour, Vegetation: Main Satellite Provision 2003-2019

NPP/ VIIRS

MODIS (Aqua)

MODIS (Terra)

Assimilation of satellite data for the environment · Summary of introduction {Use of satellite...

Documents

Transcript of Assimilation of satellite data for the environment · Summary of introduction {Use of satellite...